Meant to start a thread about this over the weekend http://82games.com/ilardi1.htm (please note up front: Yao just barely missed the amount of minutes to qualify in the main ranking) Some interesting findings.
Interesting... What struck me the most is the large amount of "error" (I'm guessing that's sort of related to a "confidence interval") in this process. Given that some combinations of players simply play a very limited number of minutes together, the "error" is pretty big even for a player that played big minutes.
I agree, given the myriad of different lineup combos & opponents, you almost need two years worth of data (actually this might be what Rosenbaum does? not sure)
It's probably hard to narrow the error much more even with 2 years worth of data, especially considering the amount of roster change within a team from year to year and the fact that each individual player doesn't play at the same level (due to injury, gain/loss in skills, aging, etc) each year. That speaks to the fact that you can't rely on +/-, even adjusted +/- exclusively when evaluationg a player (and it should be noted that Daryl Morey has said that he only uses this +/- concept as a PART of the evaluation). These +/- numbers gives you an idea as to who has been making a positive impact, you have to go to other sources (watching the tape, looking at other types of stats, etc) to figure out whether the nubmers are what they are for a good reason or as a result of fluke. I would say, however, if a player (like Shane Battier) constantly shows up as a positive in terms of +/- year in and year out and with multiple combination of teammates and on multiple teams, it is a fairly safe bet that he's doing something right on the court.
It is interesting that Shane's adjusted +/-, while good, is actually less than Rudy Gay's. But the error for players not on the extreme ends is so dominant, you probably can't draw strong conclusions from it.
Rudy Gay's +/- would not be as good on this rockets team compared to the grizz team that he is on. The way shane plays is much more effective for this rockets team
To add to this point and durvasa's point about error: It should be noted that the Grizz are a crappy crappy team. Also, adjusted +/- tries to equalize the different combinations used by one team with our without a particular player. However, I do not believe it attempts to equalize across rosters of different teams. Adjusted +/- compares how YOUR team plays with you vs. how YOUR team plays without you, not how the typical NBA team would play with or withour you. If Rudy were on a decent veteran team and the guys replacing him are competent (individually and together), his adjusted +/- may not be as high. Just like traditional box-score stats put up on bad teams may not be reliable, +/- on bad teams should also be taken with a grain of salt.
